# search

> Semantic search across the career knowledge base

- Author: Ryan Kwong
- Repository: rhkwong/high-powered-engineer
- Version: 20260122172418
- Stars: 0
- Forks: 0
- Last Updated: 2026-02-06
- Source: https://github.com/rhkwong/high-powered-engineer
- Web: https://mule.run/skillshub/@@rhkwong/high-powered-engineer~search:20260122172418

---

---
name: search
description: Semantic search across the career knowledge base
argument-hint: <query> [--results N] [--theme #tag]
allowed-tools: Bash, Read
---

# Search Skill

Search the career knowledge base using semantic similarity to find relevant advice and insights.

## Usage

```
/search <query> [--results N] [--theme #tag]
```

**Examples:**
```
/search how to get promoted to staff engineer
/search dealing with difficult manager --results 10
/search negotiating compensation --theme #compensation
/search work-life balance at senior level --results 3 --theme #career-regrets
```

## Parameters

| Parameter | Description | Default |
|-----------|-------------|---------|
| `<query>` | Natural language search query (required) | - |
| `--results N` | Number of results to return | 5 |
| `--theme #tag` | Filter by theme (e.g., `#promo-to-staff`) | None |

## Instructions

### Step 1: Parse Arguments

Extract from the user's input:
- `query`: Everything that's not a flag
- `n_results`: Value after `--results` or `-n` (default: 5)
- `theme`: Value after `--theme` or `-t` (optional)

### Step 2: Execute Search

Run the search command from the project root:

```bash
cd /Users/ryankwong/high-powered-engineer && python3 vector-search/search.py "<query>" --results <N> [--theme "<theme>"]
```

**Important:** Quote the query and theme to handle spaces properly.

### Step 3: Present Results

The search.py script outputs formatted markdown. Present the results directly to the user.

If no results are found, suggest:
- Try different keywords
- Remove theme filter if one was used
- Check if the index needs rebuilding with `/reindex`

## Error Handling

| Error | Action |
|-------|--------|
| "Index not found" | Tell user to run `/reindex` first |
| "Collection not found" | Tell user to run `/reindex --force` |
| No results | Suggest alternative queries or removing filters |

## Technical Notes

- Uses `all-MiniLM-L6-v2` embedding model (local, no API needed)
- Vector database: Chroma (stored in `vector-search/db/`)
- Relevance scores are cosine similarity (higher = more relevant)
- Each result is one H3 Topic block from the Detailed Insights sections